Parallel Traffic Management System and Its Application ... - IEEE Xplore

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 14, NO. 1, MARCH 2013

225

Parallel Traffic Management System and Its Application to the 2010 Asian Games Gang Xiong, Senior Member, IEEE, Xisong Dong, Member, IEEE, Dong Fan, Fenghua Zhu, Member, IEEE, Kunfeng Wang, and Yisheng Lv

Abstract—Field data are important for convenient daily travel of urban residents, reducing traffic congestion and accidents, pursuing a low-carbon environment-friendly sustainable development strategy, and meeting the extra peak traffic demand of large sporting events or large business activities, etc. To meet the field data demand during the 2010 Asian (Para) Games held in Guangzhou, China, based on the novel Artificial systems, Computational experiments, and Parallel execution (ACP) approach, the Parallel Traffic Management System (PtMS) was developed. It successfully helps to achieve smoothness, safety, efficiency, and reliability of public transport management during the two games, supports public traffic management and decision making, and helps enhance the public traffic management level from experience-based policy formulation and manual implementation to scientific computing-based policy formulation and implementation. The PtMS represents another new milestone in solving the management difficulty of real-world complex systems. Index Terms—Artificial systems, Computational experiments, and Parallel execution (ACP) approach, artificial transportation system (ATS), field data, parallel system, video detection.

I. I NTRODUCTION

T

HE 2010 Asian Games and Asian Para Games were the largest international events ever held in Guangzhou, China, and were an important chance to promote Guangzhou’s international friendship; regional political, economic, and cultural development level; and progress in athletic sports, amateur sports, and the sports industry. About 14 000 athletes, coaches, and sports officials, more than 7000 reporters and media personnel, more than 60 000 staff and volunteers, and about 10 million spectators and tourists from 45 countries and regions attended the Asian Games. In addition, it also involved 58 existing game facilities and 12 new sports stadiums located across Guangzhou’s metropolitan areas. A large number of contestants, spectators, and visitors could have made the existing bad transportation in Guangzhou even worse. Thus, safe and effective public traffic was essential in ensuring the success of these two games [1].

Manuscript received November 7, 2011; revised February 12, 2012 and July 10, 2012; accepted July 21, 2012. Date of publication September 10, 2012; date of current version February 25, 2013. This work was supported in part by the National Science Foundation of China under Grant 61174172, Grant 61233001, and Grant 60921061. The Associate Editor for this paper was L. Li. The authors are with State Key Laboratory of Management and Control for Complex Systems, Institute of Automation, Chinese Academy of Science (CAS), Beijing 100190, China (e-mail: gang.xiong@ia,ac.cn; xisong.dong@ ia.ac.cn (corresponding author); [email protected]; [email protected]; [email protected]; [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TITS.2012.2210883

Comparing daily traffic, transportation services during the Games had eight different characteristics. 1) Millions of passengers need high-intensity bus service. Tens of thousands of spectators and tourists and millions of residents travel between hotels, stadiums, stations, and other relevant places all over the city; more than 80% of them will use public transport tools. 2) Different passengers have different and specific traffic requirements. Different members such as officials, media reporters, athletes, sponsors, volunteers, and spectators require different and specific traffic services, such as special vehicles, special roads, a special transport fleet, specific parking spots, and specific parking lots. 3) Traffic control and security measures for the Games have an important impact on traffic. These are implemented around all venues to cause dynamic changes in traffic restrictions; therefore, all related travelers should be informed in real time. 4) The public traffic of the Asian Games is composed of many sectors, including taxi companies, bus companies, the Bus Rapid Transit (BRT) company, Metro corporations, the Guangzhou Public Security Bureau, the Communications Commission of Guangzhou Municipality, and the Command Centre for the Guangzhou Asian Games. 5) Various transport services are required, involving public transport vehicles, buses, BRT, taxies, private vehicles, rail transport, metro, fire trucks, ambulances, dedicated roads, specific cars, specific traffic lines, special signal management (signal priority), a variety of intermodal transport, parking management, traffic accident response, traffic information broadcasting, etc. 6) Fluctuation of traffic status is very difficult to predict and control. Because the number, spot, and time of traffic demand change daily as the games progress, the spatial distribution of traffic flow always changes over time, and the traffic management policy should be adaptive to those traffic distribution changes accordingly. 7) The main traffic flows are concentrated in areas such as the Asian Games venues, hotels, railway stations, and tourist attractions, which are distributed all over the city. 8) Many kinds of customers require various types of traffic information regarding the 2010 Asian Games. • Government bureaus: Traffic managers need traffic information on special vehicles, roads, field data, parking, etc. In addition, departments such as

1524-9050/$31.00 © 2012 IEEE

226


Fig. 1. Architecture of the PtMS for the 2010 Asian Games.

security, fire, medical rescue, and logistics need realtime traffic information. • Drivers need urban road and traffic status information, and route guidance. • Spectators need public traffic, parking, and transfer information, among others. • Resident travelers need traffic control and restriction information. These characteristics of public traffic services during the 2010 Asian Games posed serious management challenges, including transportation facilities, traffic control, special traffic information requirements, and traffic management and coordination. Furthermore, those complex characteristics led to many theoretic deductions and assumptions. To meet spectator and public transport demand, based on existing intelligent transportation systems (ITSs) and the novel Artificial systems, Computational experiments, and Parallel execution (ACP) approach [2] and with the help of specific traffic video detection systems and parallel traffic management systems created by the Institute of Automation, Chinese Academy of Sciences (CASIA), the Parallel Traffic Management System (PtMS) was developed to support public traffic management and decision making, which successfully helped the smoothness, safety, efficiency, and reliability of public transport management during the 2010 Asian (Para) Games [3].

The rest of this paper is organized as follows: Section II introduces the main contents and functions of the PtMS for the 2010 Asian Games. Section III presents important subsystems of the PtMS, including the Video Detection and Analysis Platform, an artificial transportation system (ATS) for the 2010 Asian Games, and the Public Traffic Evaluation System. The results and conclusions are presented in Section IV. II. PARALLEL T RAFFIC M ANAGEMENT S YSTEM FOR THE 2010 A SIAN G AMES The ACP approach is a coordinative research and systematic effort to model, analyze, and control complex systems, which are difficult or even impossible to deal with by traditional methods [2]–[8]. Basically, the ACP approach consists of three steps: 1) modeling and representation with artificial systems; 2) analysis and evaluation by computational experiments; and 3) control and management through parallel execution between actual and artificial systems [2]. Clearly, transportation systems such as large-scale urban traffic systems are suitable for the ACP approach. However, the motivation for employing the ACP approach here is mainly due to the lack of timeliness, flexibility, and effectiveness of existing transportation simulation systems [9]. The PtMS for the 2010 Asian Games (see Fig. 1) includes the actual system and the artificial system. The actual system

XIONG et al.: PTMS AND ITS APPLICATION TO THE 2010 ASIAN GAMES

includes four modules: 1) the Video Detection and Analysis Platform; 2) the Field Data Management System; 3) the Taxi Monitoring and Guidance System; and 4) the Official Traffic Website [3]. The artificial system includes another four modules: 1) the ATS; 2) the Public Traffic Evaluation System; 3) the Taxi Evaluation System; and 4) the Comprehensive Evaluation System [3]. A. Video Detection and Analysis Platform for the Asian Games The Video Detection and Analysis Platform can automatically detect real-time traffic and passenger flow at stations, main roads, and intersections and can improve the dynamic perception’s width and depth of public traffic status. With this platform, the detected traffic parameters can also be analyzed. It includes two subsystems: the traffic video detection subsystem and the crowd video detection subsystem. The traffic video detection subsystem selects traffic videos on roads leading to the 2010 Asian Games venues and other key areas. With the help of video image processing and intelligent analysis technologies, real-time traffic parameters can be collected and analyzed, including vehicle flow, time occupancy rate, average velocity, traffic density, vehicle type, queue length, etc. The crowd video detection subsystem selects videos in the 2010 Asian Games regions, with the help of video image processing and intelligent analysis technologies. Real-time pedestrian parameters at bus stations, bus stops, and other places can be acquired and analyzed, including pedestrian counting, crowd flow quantity, pedestrian density, pedestrian moving speed, etc. These analytical results are then saved into a central database for all possible applications, and quantitative data are shown on the monitor screen, together with the corresponding video stream. B. Field Data Management System for the Asian Games Based on those detected traffic parameters, the Field Data Management System provides functions such as public traffic service planning, special bus lines for the Asian Games, busscheduling management, real-time bus tracking and monitoring based on geographic information systems (GISs), plan implementation tracking and monitoring, and bus service results statistics. Based on the double-loop control principle of the ACP approach, it can provide real-time and reliable field data management, that is, its parallel system, i.e., the Public Traffic Evaluation System, can also realize adaptive optimization of public traffic management. 1) Public Traffic Service Planning: a) Venue protection planning: According to facility distribution, responsibility areas, combined with the game schedule and the distribution prediction of spectators, the Asian Games venue region, and bus numbers, bus lines can manually or automatically be decided, which also provides the basis for bus operation and monitoring. b) Special bus line planning: According to the schemes of bus lines for the Asian Games, a daily scheduling plan can be made for each line, including the time of first-run

227

and final-run vehicles, bus quantity, operation mileage, and other operation plans. 2) Special Bus Lines for the Asian Games: Basic information of special bus lines is managed, including the bus line number, station names of uplink and downlink, station code, the time of first-run and final-run vehicles, and other operation information. 3) Bus-Scheduling Management: Bus-scheduling management is applied in scheduling vehicles and drivers, which can reduce labor intensity and improve scheduling efficiency. a) Pre-defined plan management: It can edit or customize the detail operation plans of each bus line, which can be used by dispatchers for actual scheduling of the related bus lines. b) Single notice: A short message can be sent to every dedicated driver to interact with them in real time. 4) Real-Time Bus Tracking and Monitoring Based on GISs: a) Free monitoring: Vehicles can be found according to the selection criteria, such as management agency, bus line, or a dedicated parameter of any vehicle. Then, the target vehicle can be monitored in real time. b) Line monitoring: To protect the line’s service quality, any specific line can be selected for monitoring. c) Regional monitoring: To protect abundant capacity in key areas of the Asian Games, all buses in the designated areas can be selected for monitoring. d) Historical track playback: The historical track of any selected vehicle’s driving records can be played back after the time period is specified. e) Failure vehicle alarm: The operational failure buses can be found, and the alarm is sent out in real time. 5) Plan Implementation Tracking and Monitoring: The plan implementation of the venue area can be monitored by comparing the planned capacity and the actual operation capacity; the implementation status of the specified plans can be monitored according to the planned period. The plan implementation of special bus lines can be monitored. By comparing the numbers of the predefined plan and of the actual operation plan, the implementation status of the operation plan can be supervised. Violation monitoring of bus lines for the Asian Games includes time violation monitoring of first-run and final-run vehicles, quantity violation monitoring of operation buses per hour, etc. 6) Bus Service Result Evaluation: Bus service result evaluation evaluates and summarizes the implementation of the aforementioned management results. C. Taxi Monitoring and Guidance System for the Asian Games The taxi monitoring and guidance system has basic monitoring functions, advanced control functions, scheduling and guidance functions, etc. 1) Basic Monitoring: a) Global monitoring: The vehicle list can be filtered out according to distribution area, license plate number, empty

228


or manned Taxies, drivers (at work or rest), vehicle operation status (ignition or not), Taxi company name, etc. b) Vehicle location: The information of Taxi vehicles and drivers in small-scale area can be displayed. c) Vehicle tracking: With the help of a rapid, accurate, and effective information delivery channel between every taxi and the central control system, the specified vehicle’s position, speed, and status information can be tracked, and its running trajectory can be drawn out on the electronic map. d) Track playback: The history data of a specified Taxi in a defined period can be found out after setting search conditions. Its trajectory can be displayed as an evidence of passengers’ complaints, drivers’ detour behavior, passengers’ lost property, or other complaints. 2) Advanced Monitoring: a) Taxi restriction for the Asian Games: To ensure traffic smoothness in some venues or other regions (such as the railway station and the bus station) during a specific period of time during the 2010 Asian Games, all taxies can be forbidden from these places. The taxies can be monitored and warned in real time. b) Capacity monitoring: In flow distribution centers such as designated areas and official hotels or venues, taxies can be scheduled, for example, to assign specific taxies to carry specific passengers, to monitor the situation of traffic grooming delivery, to compare the actual quantity of passengers and the number of available taxies to adjust the taxies’ capacity and scheduling, and to evaluate specific scheduling implementation results of taxi companies. c) Area monitoring: Taxi operation status in the designated areas (official hotels, venues, stations, etc.), can be monitored, including the number or ratio of empty taxies to occupied taxies, the total number of taxies, etc. 3) Information Service: Information service is provided to taxi drivers by delivering short message text, such as taxischeduling information, real-time traffic situation, gas station location, and status, to taxi terminals. 4) Vehicle Scheduling and Guidance: It can scientifically formulate, evaluate and optimize the guidance program under various circumstances to provide drivers with available passenger information and real-time road conditions, to reduce empty Taxies’ time of searching passengers randomly, and to provide passengers with available Taxi information to reduce their time of waiting for an empty Taxi.

D. Official Traffic Website for the Asian Games The traffic information website provides traffic information, traffic guidance information, policy information, and weather information for all audiences, citizens, officers, and athletes. The URL of the website is http://www.gzyyjt.cn/.

E. ATS Around Tianhe Sports Center (ATS-Tianhe) ATS-Tianhe can compare several management program results under the same key performance index (KPI) and a

program’s results under different KPIs. It is composed of the ATS, Computational Experiments Platform, Statistics, and Analysis. 1) ATS: The ATS built up major road networks around the Tianhe Sports Center and simulated the different change processes of traffic flow congestion under a variety of travel modes. In addition, the results can be displayed in 3-D animation. 2) Computational Experiment Platform: Based on the ATS, the Computational Experiment Platform can realize experimental design, execution, and analysis. In addition, it can analyze and predict traffic influence under different conditions. 3) Statistics and Analysis Platform: The Statistics and Analysis Platform can compute and analyze the KPIs of the transportation targets on points (crossroads), lines (main roads), and areas (many crossroads and main roads) such as buses, the BRT, or taxies generated by the ATS. F. Public Traffic Evaluation System for the Asian Games The Public Traffic Evaluation System quantitatively evaluates man-made plans, whose evaluation objectives include the spatial–temporal saturation rate, evacuation time, etc. After the scheduling plans of public transport are inputted, the correspondent evaluation results can be calculated. By comparing the evaluation results with the actual results, decision support can be provided for decision makers to optimize the running public transport scheduling plan. After the evacuation task is finished, the quality and performance of the running public transport scheduling plan can be evaluated once again to support revising and improving the follow-up scheduling plan. The evaluation KPIs include traffic volume, average speed, traffic density, ratio of occupancy time, queue length of traffic flow, average road travel time, average travel speed, road traffic saturation, population counting, population flow rate, etc. G. Taxi Evaluation System for the Asian Games To meet the management objective, i.e., the spatial and temporal distribution ratio of empty and occupied taxies, the taxi management and scheduling plans should be evaluated beforehand, which includes passenger forecasting, taxi number, and the location and time of the taxi. Its goal is to evaluate the spatial and temporal matching ratio of all taxies and all passengers. The different results of scheduling plans can be evaluated according to the input parameters and the same evaluation KPIs; then, managers can get decision support for optimizing the current taxi-scheduling plans. H. Comprehensive Evaluation System for the Asian Games The Comprehensive Evaluation System takes the main factors into consideration at the same time, such as the general traffic, public traffic flow, weather, traffic accidents, and other traffic conditions. It can execute a comprehensive evaluation of the public traffic management plans and provide decision support, such as evaluation of the man-made decision and


Fig. 2.

229

Video detection and analysis platform.

evaluation of the special case plan. The main functions are described here. 1) Evaluation of Transport Decision Plan: The index system is constructed for the evaluation of the transport management policy. In accordance with the computational methods of the evaluation index system, evaluation indexes can be generated and stored, and evaluation results can be analyzed and compared. 2) Plan Evaluation of the Special Case: The special case plans for the 2010 Asian Games can be evaluated under different exceptional circumstances, where different traffic indicators are considered, such as the crowd evaluation time, crowd flow, crowd count, traffic saturation rate, traffic network utilization rate, average speed of traffic network, and traffic network saturation rate. 3) Three-Dimensional Display of Evaluation Results: The 3-D Display Platform of the transport policy program evaluation results for the 2010 Asian Games is developed to show the effects of different evacuation plans in 3-D scenes.

III. S OME S UBSYSTEMS OF THE PARALLEL T RAFFIC M ANAGEMENT S YSTEM FOR THE 2010 A SIAN G AMES A. Video Detection and Analysis Platform for the Asian Games The Video Detection and Analysis Platform mainly includes the traffic detection and analysis subsystem and the flow detection and analysis subsystem. After being encoded and compressed, the videos captured from the cameras are transferred to the Guangzhou Traffic Information Command Center. Then, the two subsystems will decode these signals and process each frame in real time and achieve relevant parameters, which can be simultaneously displayed on the big screen and stored in central databases (see Fig. 2).

TABLE I D ETECTION ACCURACY OF D IFFERENT T RAFFIC PARAMETERS

After integrating those information and parameters into the existing comprehensive traffic information platform, better traffic and flow management, scheduling, and evacuation programs can be achieved by the traffic management staff. 1) Traffic Video Detection and Analysis Subsystem: Through traffic videos from key areas of the Asian Games and image intelligent analysis and processing, traffic parameters and detection accuracy are shown in Table I. The function structure of the traffic video detection and analysis subsystem is shown here (see Fig. 3). • Video decoding for image acquisition is the first step, which can convert an analog video into digital signals. • Video detection for image processing: After receiving decoded images from the input device, the images are processed using embedded data detection algorithms or the traffic queue length detection algorithm. • Video compression for image compression: The video images are transmitted in real time; and therefore, the images should be compressed as much as possible in case of nonobvious image quality reduction. • Video encoding: The processed digital image is converted to analog signals, which can be output to the liquid-crystaldisplay screen for display. • Data transmission for network communication: Real-time traffic data and the compressed image can be transmitted

230


Fig. 3. Traffic video detection interface.

to the host server or other places through the wired or wireless network. 2) Crowd Video Detection and Analysis Subsystem: Using passenger crowd videos at bus stations, long-distance bus stations, and other important places, the crowd video detection and analysis subsystem can real-time count pedestrian number, crowd flow, population density, population parameters, and other statistical information. Monitor interfaces can adjust their operating parameters, such as setting the detection area, direction, alarm threshold, and alarm methods. The subsystem can also handle congestion, assaults, fighting, illegal gatherings, emergencies, and other events, such as items left behind. In addition, it can evaluate scene states based on the aforementioned pedestrian detection, tracking, and count; display their results; and put out an alarm. The subsystem is based on hierarchical modular fusion and multistate switching to achieve pedestrian detection, tracking, and statistical count. Its overall operating flowchart is shown in Fig. 4, and the subsystem interface is shown in Fig. 5. The subsystem uses a modular approach to decompose the complex problem into a comparatively easy subprocess. B. ATS-Tianhe The ATS is a new method to research complex traffic systems [10]–[18]. 1) ATS’s Architecture: The ATS has a four-level architecture, as shown in Fig. 6. The fundamental component level is its foundation. It takes the predefined basic components as its program components. Through the interaction of components and external interfaces, the system’s flexibility and scalability can be fundamentally guaranteed. The related components include urban planning models, vehicle motion models, population models, staff mental models, road signal control models, etc., which can model the traffic behaviors of passengers and vehicles through business process orchestration of component services [10]. The data and knowledge level is the core of the ATS. Its control and management models are denoted as the model

algorithm library and expert knowledge base [10]–[12]. Based on these models, experimental scenarios can be provided [19]. The computational experimental level is the function realization module of the ATS, which can build specific traffic scenarios by using data and knowledge from the data and knowledge level. It can achieve evaluation and decision support of actual traffic control and management programs by experiment design, execution, and analysis [19]–[21]. The parallel execution level is a layer of function display and user interface, which can execute control and management of the actual system through information exchange and feedback. Through a variety of front-end analysis tools, interactive communication from the actual transportation system, and forecasting, analysis and evaluation results from the ATS can be used to guide the operation of the actual transportation system. The transport plans can be evaluated in advance. In addition, scientific, accurate, and timely information and knowledge can be provided for decision makers. 2) Construction Process of ATS: Step 1: Construct the fundamental component level. Serviceoriented architecture (or cloud computing) platforms with large-scale distributed data storage and distributed computing capabilities are built. A multiagent environment can be established according to the Foundation for Intelligent Physical Agent Specifications; Agent Management System, Distributed Directory Facilitator, Agent Communication Channel, and other multiagent platform components can be developed to achieve the agent within the platform lifecycle services, yellow page services, and messaging services. Finally, combining the specific applications of the experimental platform, application-domain ontology can be built up to achieve internal semantic interoperability among interagencies. Step 2: Construct the data and knowledge level. The agentbased model library of the participants, environment, rules, and mechanisms of application and demonstration of specific areas can be constructed. Then, a scenario library can be built on real-time monitoring data. XML-based language can be used to design a unified formal representation to describe the knowledge of experimental platform, and a domain knowledge library can semiautomatically be built up by using machine learning and natural language processing technology to achieve storage, expression, and reasoning within the platform. Step 3: Construct the computing platform. Based on the requirements of the parallel control and management system, the computing platform can be constructed, including the massive data storage and exchange system, high-performance computers, large-scale networking environments, operating systems, infrastructure software, application development tools, etc. Step 4: Construct the computational experimental level. It can accept scenes input by the end user or scenes in the scene library and their instantiate interactive


Fig. 4.

Fig. 5.

231

Flowchart of the crowd video detection and analysis system.

Crowd video detection and analysis system.

mechanisms and management rules, which can be passed to the event-driven engine to complete the simulation experiments. The event-driven engine can be constructed on discrete event simulation technology, which can dynamically simulate the process of interaction and communication between agents in the experimental scenes. It can simulate a specific moment and time when the platform is running a simulation clock and can store, analyze, and determine the course of the triggered relations between discrete events in chronological order. The experimental process can be driven and simulated by the advance of the simulation clock and the handling of discrete events. Finally, the algorithm analysis tools need to be developed for the experimental platform, which can be used as modules and components. The strategy learning and optimization algorithms, qualitative and quantitative evaluation algorithms,

and various proprietary algorithms need to be developed, which can dynamically analyze, evaluate, and optimize the process of computational experiments and their results, and update the knowledge base in real time. Step 5: Construct the parallel execution level. Based on the intelligent detection system and sensor networks, traffic, pedestrian flow, and other relevant data can be monitored, collected, and integrated in real time, and experimental scenes can be generated and adjusted according to real-time monitoring data. Second, complete software libraries and high-level application protocols can be designed, which can provide service for the interfaces between experimental platforms and end users, and end users can easily manage and configure the platforms and the agency operation. The unsafe factors of the experimental scenes can be monitored and evaluated in real time to achieve passive query, active risk evaluation, and early warning of event safety. At the same time, by computational experiments and feedback control, semiautomatic evaluation and optimization can be achieved, and optimal decision can be generated. Finally, the dynamic visual human–computer interface can be designed, which can show all computational experiments and their interactive control process with text, graphs, charts, and other forms. 3) Construction of ATS-Tianhe: ATS-Tianhe can be established by integrating GIS information, traffic maps, population distribution, weather data, scheduling plans of the Asian Games, participant distribution, and other data. a) Surrounding environment of ATS-Tianhe: The basic traffic surrounding environment of the Tianhe Sport Center needs traffic network, public traffic lines, testing equipment, regulations of public activity, and distribution of private cars.

232


Fig. 6. Public traffic evaluation system.

Its population distribution includes the population distribution around the Tianhe Sport Center, sex ratio, age division and ratio, population growth type, travel habits, origin–destination matrix, travel speed distribution, and other information. Its vehicle information is constructed on traffic data around the Tianhe Sport Center, including traffic distribution, number of cars, private car ownership, vehicle type distribution, etc. b) Traffic management of surrounding environment: It can achieve management of traffic signal control, public traffic scheduling, public traffic management, etc. Traffic signal control management includes signal control modules and control methods (phase and time) of every traffic signal controller at the main intersections. Public traffic scheduling management includes the time, number, type, capacity, frequency, and stations of every public traffic line. Public traffic management includes velocity limitation, traffic restriction according to vehicle type and plate number, and other traffic management regulations. c) Environmental influence on traffic: Weather influence on traffic includes weather, accidents, game arrangements, etc. It can be constructed as different typical weather models according to climate data of Guangzhou. Every model can represent a specific type of weather. Accident influences can be constructed as different typical accident models, which can model the traffic influence of regular accidents and unconventional emergencies. Accident models can be divided into different types of accidents. Everyone can simulate a kind of accident according to the general impact and duration time. Game arrangements can be constructed as models of every game held at the Tianhe Sport Center, including the beginning and ending time, game time, game type, number of seats, participant number and distribution, etc. d) ATS-Tianhe: ATS-Tianhe is composed of 243 roads, 76 crossroads, about 100 000 people, and 320 urban sites (which consists of 147 residential communities, 67 office communities, 35 schools, 18 shopping centers, 10 recreation centers, 14 restaurants, 10 emergency evacuation sites, and 19 other sites). e) Experiment and evaluation based on ATS-Tianhe: Based on ATS-Tianhe, the experiments of traffic evacuation plans can be designed and simulated to show an evaluation report by analyzing experimental results, according to the

traffic evacuation requirements and objectives. Different traffic scenarios can be built in the traffic evacuation experiments by selecting variable artificial transportation models, including not only traffic scenarios under normal conditions but traffic scenarios under abnormal environments such as traffic accidents, terrorism attacks, and adverse weather conditions as well. During the evaluation process, a variety of traffic parameters in the designed traffic scenarios can be counted, summarized, mined, extracted, and integrated to get the parameter reports of different traffic scenarios in different modes. With the help of ATS-Tianhe, we were able to compute different possible results of several traffic management programs using the same KPI, and the results of the same programs using different KPIs can be computed and analyzed. Thus, we could evaluate those possible results under different traffic situations to determine how to improve them, and therefore, daily operation of the actual transportation system could also be improved. C. Public Traffic Evaluation System for the Asian Games The Public Traffic Evaluation System is constructed by combining the parallel evaluation system with the evaluation index hierarchy (see Fig. 7). 1) Parallel Evaluation System: The parallel evaluation system can carry out computational experiments of traffic evacuation based on the ATS. Considering the people, vehicles, roads, environment, traffic demand, decision-making programs, vehicle-scheduling program, and traffic-handling program in special circumstances, it can evaluate the implementation of the evacuation program and give the optimization results from the perspective of global simulation. It can also optimize the actual system, prevent accidents in normal circumstances, and carry out emergency management to improve the management capabilities in abnormal circumstances. The ATS receives a variety of control and management strategies and the current system state (initial point) from the actual system, verifies feasibility and implementation effect, and evaluates the effectiveness of the program (decision support). The parallel evaluation system can run offline or online. In offline mode, the parallel system can provide a simulation environment of learning and training, where traffic managers


Fig. 7.

233

ATS architecture.

can “experiment” on a program before system implementation. In online mode, traffic managers can quickly and effectively grasp traffic running states and respond. 2) Evaluation Index Hierarchy: The traffic evaluation index hierarchy can be used for the Public Traffic Evaluation System, Taxi Evaluation System, and Comprehensive Evaluation System. Several evaluation index hierarchies can be constructed, which can be used for the ATS and the actual transportation systems. The evaluation indexes of the ATS are divided into four types according to different evaluation objectives. a) Evaluation index for sections. During the operation of ATS-Tianhe, all sections can be counted and analyzed for every cycle time, and the corresponding values of all statistical indicators can be generated. There are a total of 35 indicators, including the traffic flow rate; total and average mileage and delay time; average flow rate; average velocity; average road coverage; exhaust emissions of HC, CO, and NOx ; total, average, and maximum Uturn, turn-around, turn-left, straight, and turn-right queue length; and so on. b) Evaluation indexes for intersections. Evaluation indexes for intersections take intersections and their surrounding as a whole. Like evaluation indexes based on sections, these include parking number, average parking time, delay time, section saturation, green signal utilization ratio, etc. c) Evaluation indexes for traffic networks. Evaluation indexes for traffic networks take all vehicles on the road as a whole. Like evaluation indexes based on the link method, these include vehicle number; average velocity; exhaust emissions of HC, CO, and NOx ; etc. d) Evaluation indexes for activities types. Different from the aforementioned three types of indexes, these indexes

TABLE II S AATY ’ S 9-S CALE M ETHOD

are not calculated according to the survey cycle but are counted once after one running time. They include the average delay time, average parking number, average travel time, etc. Based on preceding discussion, comprehensive evaluation of the system-level structure is given as follows: a) Construction of the traffic evaluation index hierarchy: Different traffic evaluation index hierarchies that can reflect different evaluation levels of the public transport system can be constructed. For the index hierarchy, after selecting the indexes, the relative importance degree can be selected according to expert experimental analysis to form the following T-Ck matrix: comparison matrix W, where Ci and Cj represent the indexes, and Wij represents the importance degree of Ci , compared to Cj . In W, Cj in the columns are taken as the denominator, and we take the nine-scale method proposed by American operational researcher A. L. Saaty. That is, natural numbers 1–9 and their reciprocals denote the relative importance degree (see Table II). For W, calculating its eigenvectors, λmax is the largest one, and P is the orthonormal vector according to λmax .

234


Component Pi of P is the single sort weight corresponding to indicators Ci .

in index hierarchy. Accumulating their results, the evaluation coefficients in index hierarchy can be obtained.

1) Orthonormalize each column of W, bij = (wij / n i=1 wij ). 2) Calculate the sum of every row of each orthonomalized column of W: Zi = nj=1 bij . 3) Z = [Z1 , Z2 , . . . , Zn ]T can be orthonormalized as pi = (zi / ni=1 zi ). Then, the eigenvectors of W is P = [P1 , P2 , . . . , Pn ]T . 4) Calculate the largest eigenvalue of W: λmax = n i=1 ((W P )i /nPi ), where (W P )i is the ith component of W P .

1) Collect the value sequence of each sample period x1 , x2 , . . . , xi . 2) Set the direct and reverse relation of an index and maximum M and minimum m. 3) For each xi , calculate its precalculation ri . 4) For each ri , multiply eigenvector pi in its index hierarchy to get the index coefficient Ei = ri ∗ pi . 5) Accumulate all index coefficients to get index coefficient E, i.e., E = Ei .

W is based on the importance degree from subjective judgments of experts. When experts are analyzing, the characteristic Wij = Wik /Wjk is difficult to achieve; therefore, to alleviate the burden of expert analysis, inconsistent judgments are allowed to some extent. Consistency index CI = (λmax − n/n − 1) can control the extent of the matrix. When CI is larger, the deviation extent of W from the consistency is larger. When CI is smaller, W is closer to consistency. Define RI as the average value of CI. The random consistency index ratio CR is the ratio of CI and RI, i.e., CR = CI/RI. Consistency check: When CR < 0.1, W has an acceptable consistency. When CR ≥ 0.1, the matrix should be adjusted to meet CR < 0.1 and should have satisfactory consistency. b) Application of traffic evaluation index hierarchy: The regular evaluation system can be established by traffic evaluation index hierarchy. When the corresponding index data of the ATS are collected according to the evaluation system, evaluation indexes can be calculated in accordance with the evaluation algorithm, which can reflect the running effect of the ATS. The evaluation algorithm has the following process: under multiindex evaluation conditions, the values can be reflected in the evaluation of each single index, and their positive and negative indicators need to be orthonormalized. The assumptions are given here. 1) Each index value is in a directly or inversely proportional simple linear relationship. 2) The best result of evaluation is the situation where the index has maximal or minimal value. Indicator data reflect forward or reverse goodness or badness of evaluation conclusions. For an indicator of positive evaluation, its precalculation is 1, x i ≥ Mi xi −mi , mi < x i < M i ri = M i −mi 0, x i ≤ mi . Similarly, reverse evaluation of the precalculation is 1, x i ≥ Mi Mi −xi , mi < x i < M i ri = M i −mi 0, x i ≤ mi . Experimental index data of direct or reverse orthonormalizaion are calculated according to their eigenvector matrix

IV. R ESULTS AND C ONCLUSION With the help of existing ITSs, traffic video detection novel technology and parallel traffic management systems are introduced to build up the PtMS especially for the 2010 Asian Games. The PtMS successfully helped to enhance the field data management level from experience-based policy formulation and manual implementation to scientific-computing-based policy formulation and implementation. The actual data of October 11, 2010 (Monday, before the PtMS was used) and October 18, 2010 (Monday, after the PtMS was used) of the comprehensive public traffic evaluation have been taken as a case study. Thus, the PtMS increases the average speed of the road network from 17.96 to 20.10 km/h. After the Asian Games, the PtMS was successfully modified and applied for BRT management. In addition, the PtMS can also be applied for public transport management during the Chinese Spring Festival and other major holidays, the China Import and Export Fair (Canton Fair), and other large commercial activities. The successful implementation of the PtMS can continuously provide field data and valuable experiences to test and evaluate the effectiveness of the ACP approach for solving the management difficulty of real-world complex systems. R EFERENCES [1] G. Xiong, K. F. Wang, F. H. Zhu, C. Cheng, X. J. An, and Z. D. Xie, “Parallel traffic management for 2010 Asian Games,” IEEE Intell. Syst., vol. 25, no. 3, pp. 81–85, May/Jun. 2010. [2] F. Y. Wang, “Artificial societies, computational experiments, and parallel systems: An investigation on computational theory of complex socialeconomic systems,” Complex Syst. Complexity Sci., vol. 1, no. 4, pp. 25– 35, Apr. 2004. [3] G. Xiong, S. Liu, X. S. Dong, F. H. Zhu, B. Hu, D. Fan, and Z. Zhang, “Parallel traffic management system helps 16th Asian games,” IEEE Intell. Syst., vol. 27, no. 3, pp. 74–78, May/Jun. 2012. [4] F. Y. Wang, “Toward a paradigm shift in social computing: The ACP approach,” IEEE Intell. Syst., vol. 22, no. 5, pp. 65–67, Sep./Oct. 2007. [5] F. Y. Wang and S. Tang, “Artificial societies for integrated and sustainable development of metropolitan systems,” IEEE Intell. Syst., vol. 19, no. 4, pp. 82–87, Jul./Aug. 2004. [6] B. Ning, T. Tang, H. R. Dong, D. Wen, D. R. Liu, S. G. Gao, and J. Wang, “An introduction to parallel control and management for highspeed railway systems,” IEEE Trans. Intell. Transp. Syst., vol. 12, no. 4, pp. 1473–1483, Dec. 2011. [7] B. Ning, H. R. Dong, D. Wen, L. F. Li, and C. C. Cheng, “ACP-based control and management of urban rail transportation systems,” IEEE Intell. Syst., vol. 26, no. 2, pp. 84–88, Mar./Apr. 2011. [8] F. Y. Wang, “Parallel control and management for intelligent transportation systems: Concepts, architectures, and applications,” IEEE Trans. Intell. Transp. Syst., vol. 11, no. 3, pp. 630–638, Sep. 2010.


[9] F. H. Zhu, F. Y. Wang, R. M. Li, Y. S. Lv, and S. H. Chen, “Modeling and analyzing transportation systems based on ACP approach,” in Proc. 14th Int. IEEE Conf. Intell. Transp. Syst., 2011, pp. 2136–2141. [10] J. Y. Li, S. M. Tang, X. Q. Wang, W. Duan, and F.-Y. Wang, “Growing artificial transportation systems: A rule-based iterative design process,” IEEE Trans. Intell. Transp. Syst., vol. 12, no. 2, pp. 322–332, Jun. 2011. [11] F. Y. Wang and S. M. Tang, “A framework for artificial transportation systems: From computer simulations to computational experiments,” in Proc. IEEE Conf. Intell. Transp. Syst., 2005, pp. 1130–1134. [12] J. Y. Li, S. M. Tang, X. Q. Wang, and F. Y. Wang, “A software architecture for artificial transportation systems—Principles and framework,” in Proc. 10th IEEE Int. Conf. Intell. Transp. Syst., 2007, pp. 229–234. [13] J. Y. Li, S. M. Tang, and F. Y. Wang, “An investigation on ATS from the perspective of complex systems,” in Proc. 11th IEEE Int. Conf. Intell. Transp. Syst., 2008, pp. 20–24. [14] L. F. Li, H. Zhang, X. F. Wang, W. Lu, and Z. P. Mu, “Urban transit coordination using an artificial transportation system,” IEEE Trans. Intell. Transp. Syst., vol. 12, no. 2, pp. 374–383, Jun. 2011. [15] F. H. Zhu, Z. X. Wang, F.-Y. Wang, and S. M. Tang, “Modeling interactions in artificial transportation systems using Petri net,” in Proc. IEEE Conf. Intell. Transp. Syst., 2006, pp. 1131–1136. [16] Q. H. Miao, F. H. Zhu, Y. S. Lv, C. J. Cheng, C. Chen, and X. G. Qiu, “A game-engine-based platform for modeling and computing artificial transportation systems,” IEEE Trans. Intell. Transp. Syst., vol. 12, no. 2, pp. 343–353, Jun. 2011. [17] F. He, Q. H. Miao, Y. T. Li, F. Y. Wang, and S. M. Tang, “Modeling and analysis of artificial transportation system based on multi-agent technology,” in Proc. IEEE Conf. Intell. Transp. Syst., 2006, pp. 1120–1124. [18] J. Z. Liu, R. M. Li, F. H. Zhu, and H. R. Dong, “Modeling on agent-based artificial field data,” in Proc. 30th Chin. Control Conf., 2011, pp. 5596– 5600. [19] Y. Ou, S. M. Tang, and F. Y. Wang, “Computational experiments for studying impacts of land use on traffic systems,” in Proc. 13th IEEE Conf. Intell. Transp. Syst., 2010, pp. 1813–1818. [20] K. Wang and Z. Shen, “Artificial societies and GPU-based cloud computing for intelligent transportation management,” IEEE Intell. Syst., vol. 26, no. 4, pp. 22–28, Jul./Aug. 2011. [21] N. Zhang, F. Y. Wang, F. Zhu, D. Zhao, and S. Tang, “DynaCAS: Computational experiments and decision support for ITS,” IEEE Intell. Syst., vol. 23, no. 6, pp. 19–23, Nov./Dec. 2008. [22] C. Oh, J. S. Oh, and S. G. Ritchie, “Real-time hazardous traffic condition warning system: Framework and evaluation,” IEEE Trans. Intell. Transp. Syst., vol. 6, no. 3, pp. 265–272, Sep. 2005.

235

Xisong Dong (M’12) received the B.Sc. degree in applied mathematics and the Ph.D. degree in control theory and control engineering from the University of Science and Technology Beijing, Beijing, China, in 2001 and 2007, respectively. From 2007 to 2010, he was a Postdoctor with the Center of Information Security, Beijing University of Posts and Communications. He is currently an Assistant Professor with the State Key Laboratory of Management and Control for Complex Systems, Beijing Engineering Research Center for Intelligent Systems and Technology, Institute of Automation, Chinese Academy of Sciences. He is the author of more than 40 journal or conference proceeding papers. His research interests include the modeling and analysis of complex systems and the modeling and optimization of transportation systems and complex grids.

Dong Fan received the B.S. degree in computer science and applications from Yantai University, Yantai, China. He is currently working toward the M.S. degree with the College of Computing and Communication Engineering, Graduate University of Chinese Academy of Sciences, Beijing, China. His research interests include enterprise intelligence, information solutions of complex systems, system modeling and schema designing, and software engineering management and control.

Fenghua Zhu (M’10) received the Ph.D. degree in control theory and control engineering from the Chinese Academy of Sciences (CAS), Beijing, China, in 2008. He is currently an Associate Researcher Scientist with the State Key Laboratory of Management and Control for Complex Systems, Beijing Engineering Research Center for Intelligent Systems and Technology, Institute of Automation, Chinese Academy of Sciences. His research interests include artificial transportation systems and parallel transportation management systems.

Gang Xiong (SM’10) received the B.S. and M.S. degrees from Xi’an University of Science and Technology, Xi’an, China, in 1991 and 1994, respectively, and the Ph.D. degrees in control science and engineering from Shanghai Jiao Tong University, Shanghai, China, in 1996. From 1996 to 1998, he was a Postdoctor and Associate Scientist with the Institute of Industrial Process Control, Zhejiang University. From 1998 to 2001, he was a Senior Research Fellow with the Automation and Control Institute, Tampere University of Technology, Tampere, Finland. From 2001 to 2007, he was a Specialist and Project Manager with Nokia Corporation, Finland. In 2007, he was a Senior Consultant and Team Leader with Accenture and Chevron, USA. In 2008, he was the Deputy Director of the Informatization Office, Chinese Academy of Science (CAS), Beijing, China. In 2009, he started as a Research Scientist with the State Key Laboratory of Management and Control for Complex Systems, Beijing Engineering Research Center for Intelligent Systems and Technology, Institute of Automation, CAS. He has successfully finished more than 30 academic or technical projects. He is the author of about 160 refereed journal and conference proceeding papers. His research interests include parallel control and management, modeling and optimization of complex systems, and intelligent transportation systems. Dr. Xiong is a member of the Institute for Operations Research and Management Sciences, the International Council on Systems Engineering, and the Chinese Association of Automation.

Kunfeng Wang received the B.Eng. degree from Beihang University, Beijing, China, in 2003 and the Ph.D. degree in control theory and control engineering from the Chinese Academy of Sciences (CAS), Beijing, in 2008. Since July 2008, he has been an Assistant Professor with the Institute of Automation, CAS. His research interests include video image processing, machine learning, and traffic modeling and control.

Yisheng Lv received the Ph.D. degree in control theory and control engineering from the Chinese Academy of Sciences (CAS), Beijing, China, in 2010. He is currently an Assistant Professor with the State Key Laboratory of Management and Control for Complex Systems, Beijing Engineering Research Center for Intelligent Systems and Technology, Institute of Automation, CAS. His research interests include intelligent transportation systems and emergency traffic management.